Many web hosting shoppers are weary of VPS (Virtual Private Server) solutions that are touted as offering the performance of a dedicated server but as a software instance. High end Virtual Private Servers can usually outperform an entry level dedicated server but everything is relative.

I decided to do some benchmarks today to identify the performance indicators and compare it to one of my traditional desktop systems. Three benchmark runs were done, with the average results chosen of the three benchmarks.

The results below were produced by an AMD Athlon X2 5200+ with 2 Gigabytes of memory and a run of the mill SATA 7200 RPM hard disk. It had Ubuntu 32 Bit 10.04 installed:

BYTE UNIX Benchmarks (Version 4.1-wht.2)
System — Linux 2.6.32-14-generic #20-Ubuntu SMP Sat Feb 20 05:38:50 UTC 2010 i686 GNU/Linux
/dev/sda7             77813696  63782568  10078340  87% /

Start Benchmark Run: Tue Jun 15 13:39:58 CAT 2010
13:39:58 up 15:43,  2 users,  load average: 0.02, 0.06, 0.01

End Benchmark Run: Tue Jun 15 13:50:31 CAT 2010
13:50:31 up 15:53,  2 users,  load average: 18.72, 7.08, 2.99

INDEX VALUES
TEST                                        BASELINE     RESULT      INDEX

Dhrystone 2 using register variables        376783.7 11393094.3      302.4
Double-Precision Whetstone                      83.1     1269.0      152.7
Execl Throughput                               188.3     7209.1      382.9
File Copy 1024 bufsize 2000 maxblocks         2672.0   131291.0      491.4
File Copy 256 bufsize 500 maxblocks           1077.0    46040.0      427.5
File Read 4096 bufsize 8000 maxblocks        15382.0   772334.0      502.1
Pipe-based Context Switching                 15448.6   411501.4      266.4
Pipe Throughput                             111814.6  1580693.0      141.4
Process Creation                               569.3    20060.0      352.4
Shell Scripts (8 concurrent)                    44.8      900.8      201.1
System Call Overhead                        114433.5  2763156.4      241.5
=========
FINAL SCORE                                                          289.6

The results below are from our Silver VPS line, featuring Virtuozzo virtualisation. The VM was installed with CentOS 5 64 bit with cPanel and sports 512 Megabytes of memory.

BYTE UNIX Benchmarks (Version 4.1-wht.2)
System — Linux 2.6.18-028stab067.4 #1 SMP Thu Jan 14 17:06:11 MSK 2010 i686 i686 i386 GNU/Linux
/dev/vzfs             31457280   9656548  21800732  31% /

Start Benchmark Run: Tue Jun 15 09:58:08 EDT 2010
09:58:08 up 23 days, 10:43,  2 users,  load average: 0.02, 0.03, 0.00

End Benchmark Run: Tue Jun 15 10:08:13 EDT 2010
10:08:13 up 23 days, 10:53,  2 users,  load average: 19.44, 7.53, 3.12

INDEX VALUES
TEST                                        BASELINE     RESULT      INDEX

Dhrystone 2 using register variables        376783.7 21662200.7      574.9
Double-Precision Whetstone                      83.1     1217.9      146.6
Execl Throughput                               188.3     7482.7      397.4
File Copy 1024 bufsize 2000 maxblocks         2672.0   245885.0      920.2
File Copy 256 bufsize 500 maxblocks           1077.0    68374.0      634.9
File Read 4096 bufsize 8000 maxblocks        15382.0  2397244.0     1558.5
Pipe-based Context Switching                 15448.6   256611.7      166.1
Pipe Throughput                             111814.6  3229997.3      288.9
Process Creation                               569.3    25387.6      445.9
Shell Scripts (8 concurrent)                    44.8     2504.2      559.0
System Call Overhead                        114433.5  3037263.6      265.4
=========
FINAL SCORE                                                          431.6

You may be wondering why the VPS performed so well. The result can be explained because of the Equal – Share CPU setup and SAS 15 000 RPM RAID-10 disk configuration. Each VPS node consists of at least 16 CPU cores of 2.2 GHz and you can attain very high speeds with these configured for Equal Share. When you couple that with the I/O performance of the disk configuration, you can see why there is a huge difference in the score, in spite of available memory.

Want to try benchmarking your machine? Feel free to download the modified BYTE Magazine benchmark here

To install:

# tar -xzvf unixbench-*-wht.tar.gz
# cd unixbench-*-wht-2 ; make
# ./Run

Let me know if there are any specific benchmarks you want to see. The next batch of benchmarks will comprise of Xen based virtual machines, as well as more multi – threading intensive tests.